arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2603.10032cs.ARcs.AIcs.LG
收藏

HTM-EAR: 在饱和条件下结合混合路由的重要性保持分层内存

HTM-EAR: Importance-Preserving Tiered Memory with Hybrid Routing under Saturation

Shubham Kumar Singh

AI总结:

HTM-EAR通过混合路由和重要性感知淘汰,在内存饱和条件下有效管理事实,保持查询精度并控制历史遗忘。

摘要:

在长时间运行的智能体中,内存限制要求对积累的事实进行结构化管理,同时在有限的上下文限制下保留关键信息。我们介绍了HTM-EAR,一种分层分层内存子系统,它将基于HNSW的工作内存(L1)与存档存储(L2)相结合,结合了重要性感知的淘汰和混合路由。当L1达到容量时,项目通过重要性和使用量的加权分数进行淘汰。查询首先在L1中解决;如果相似性或实体覆盖不足,检索会回退到L2,并使用交叉编码器对候选者重新排序。我们使用合成流在五个随机种子和真实BGL系统日志上评估该系统,持续饱和(15,000个事实;L1容量500;L2容量5000)。消融研究将完整系统与无交叉编码器重新排序、无路由门、LRU淘汰和无界内存的oracle进行比较。在饱和条件下,完整模型保持了主动查询的精度(MRR = 1.000),同时允许对过时历史进行受控遗忘,接近oracle主动性能(0.997 ± 0.003)。相比之下,LRU最小化延迟(21.1 ms)但永久删除2416个关键事实。在BGL日志上,完整系统实现了MRR 0.336,接近oracle(0.370),而LRU则降至0.069。代码可在:https://github.com/shubham-61291/HTM-EAR上公开获取。

英文摘要:

Memory constraints in long-running agents require structured management of accumulated facts while preserving essential information under bounded context limits. We introduce HTM-EAR, a hierarchical tiered memory substrate that integrates HNSW-based working memory (L1) with archival storage (L2), combining importance-aware eviction and hybrid routing. When L1 reaches capacity, items are evicted using a weighted score of importance and usage. Queries are first resolved in L1; if similarity or entity coverage is insufficient, retrieval falls back to L2, and candidates are re-ranked using a cross-encoder. We evaluate the system under sustained saturation (15,000 facts; L1 capacity 500; L2 capacity 5000) using synthetic streams across five random seeds and real BGL system logs. Ablation studies compare the full system against variants without cross-encoder re-ranking, without routing gates, with LRU eviction, and an oracle with unbounded memory. Under saturation, the full model preserves active-query precision (MRR = 1.000) while enabling controlled forgetting of stale history, approaching oracle active performance (0.997 +/- 0.003). In contrast, LRU minimizes latency (21.1 ms) but permanently evicts 2416 essential facts. On BGL logs, the full system achieves MRR 0.336, close to the oracle (0.370), while LRU drops to 0.069. Code is publicly available at: https://github.com/shubham-61291/HTM-EAR

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑